An Overview Of Machine Learning Definition and Types

The term AI and Machine learning get more popular recently because of the huge amount of daily data produced on the internet, besides the high computational power of computer systems.

Every time you interacts with a platform such as social networks or e-commerce website, your data are stored on the cloud. Imagine millions of internet users around the globe using tons of websites every day, How much data will be there?
By 2025, 200+ zettabytes of data are estimated to be stored on the cloud.[1]

From education, farming, manufacturing to healthcare, More Industries are interested in integrating AI in their work process and creating more smart technologies that ease their tasks.

In this article, we will take a look to comprehend the exact meaning of machine learning and its major types with realistic use cases.

Machine Learning definition :

Machine learning is a subfield of Artificial Intelligence that provides systems the capacity to learn and improve automatically without being previously programmed. Machine learning focuses especially on building systems that can learn from different types of data and make autonomous decisions without any human intervention.

How does a machine learn?

The learning process of a machine starts by observing and looking for patterns using specific algorithms among the data loads used such as numbers, text, images, videos, and speech. Finding more patterns will help the system to make decisions in the future based on the examples provided. 

Supervised machine learning 

Supervised machine learning is the most used type of machine learning algorithm.
It uses labeled data to train an algorithm that will learn the mapping function from the input ( labeled data ) to the output ( the prediction ).

The process of an algorithm learning from data that are labeled with the correct answer is similar to having a teacher supervising the process. That’s why it is called supervised learning.
the algorithm can achieve a good level of performance to predict the output for data when the mapping function between output variables and output variables is approximated well.
In other words, when the loss function is minimal, we can expect a high accuracy of our model to predict the right answer. 
Common examples of supervised learning: classifying spam emails.

Supervised problems are grouped into two types:
Classification: Predicting a discrete value such as Male, Female, True, False…
Regression: Predicting a continuous value such as Price, Weight, Salary…

Let’s understand supervised learning through a practical example, we want to predict the salary of an employee based on their experience. we will get the following chart by drawing the best line that fits the points. the result function will let us predict new data never seen by the model (this is a regression problem).

Famous supervised learning algorithms:

• Linear Regression
• Logistic Regression
• Decision Tree
• Support Vector Machine
• K-nerest naighbor
• Random Forest
• Naïve Bayes
• Ensemble methods 

Unsupervised learning

Unsupervised learning discovers the hidden patterns and the underlying structure from unlabeled data. it is often used in data analysis to understand the distribution.Unlike supervised learning, the algorithm learns on its own without a supervisor and there is no correct answer.
Unsupervised learning problems can be grouped into Clustering and association.

Popular Unsupervised learning algorithms:

• Partitional Clustering: k-mean
• Hierarchical clustering

Semi-Supervised Learning

In Semi-supervised Learning, some input data are labeled and the rest is unlabeled.
Many techniques are used in the case of semi-supervised learning. For example, you can use Unsupervised learning to discover the structure of the input data or use supervised learning to predict the unlabeled data and turn the problem into a supervised learning problem.

[1] https://www.statista.com/statistics/871513/worldwide-data-created/